MioFace ID — Live Demo

⬆ Upload image

Drop an image here to identify

📷

Press ▶ Start camera to begin.
Your camera stays on this device; frames are matched against the gallery.

Capture quality —

Start the camera for guided capture.

⬆ Add from disk

Live tuning

Identify threshold 0.45 Below this, a face is reported Unknown. Higher = stricter. Top-K candidates 5 Max faces recognised / frame 3 In a crowd, faces are recognised a few per frame (round-robin) and tracked — boxes stay smooth, labels persist. Quality gate (reject tiny / blurry faces) ROI send (crop around faces — watch KB on the live badge)

Engine

Recognition analytics

How cleanly the engine separates identities, and how 1:N accuracy holds as the gallery size N grows.

Genuine vs impostor similarity — same vs different people

ROC curve — true-accept vs false-accept

FAR / FRR vs threshold — choose an operating point

CMC — rank-N identification

Rank-1 accuracy vs gallery size N — the scaling story (log-x)

Embedding map — each point is a face, coloured by source

Images per identity

Identities by source

Register a person

Live tuning

Engine

Recognition analytics